DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 3/10/2026 has been entered.
Priority
Applicant claims no benefit of any prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, or 365(c). Priority is given to US application 17/453,705 for the initial filing date 11/5/2021.
Status of Claims
Claims 7, 16 and 21 are cancelled.
Claims 1-6, 8-15, 17-20 and 22-23 are pending and examined on the merits.
Withdrawn Rejections/Objections
The rejections to claims 1-6, 8-15 and 17-22 are rejected under 35 U.S.C. 112(a)
in the Office action mailed 08 January 2026 are withdrawn in view of claim amendments filed 3/10/2026
Claim Rejections - 35 USC § 112
This is a newly installed rejection. Necessitated by claim amendments.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-6, 8-9, 14, 18, 20 and 22-23 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites “storing, by the one or more processors, the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer” at the last step. It is not clear why storing the frequent k-mer of viral replication origin k-mers based on (a) the output value and (b) and a character length of the frequent k-mer. No other claim steps used the frequent k-mer by (a) the output value and (b) and a character length of the frequent k-mer.
The (a) “output value” is interpreted as the frequency of the k-mer (the value for the key-value pair of any hash data structure). “Based on (a) the output value” is interpreted as only a subset of the Hash component wherein the frequency is bigger than a threshold is stored. The “character length of the frequent k-mer” is interpreted as the string length of the frequent k-mers. However, it is not clear who benefited from, and who ever used the stored k-mers sorted by the k-mer length.
Claim 14 recites “identifying a hypothesis k-mer of the plurality of frequent k-mers” at the first step. This limitation is unclear because a k-mer cannot be identified as part of the plurality before determining the membership unless the process is clarified.
Claim 18 recites “a demographic frequency measure” by “a location-wide designation”. The relationship between location and demographic designation is unclear, because there is usually no simple 1:1 correlation between a location and a demographic designation.
Claim 20 recites the limitation "the first genomic sequence set" in the 4th line. There is insufficient antecedent basis for this limitation in the claim. Claim 19 recites “a genomic sequence set”.
Claim Rejections - 35 USC § 101
The current rejection is modified from the previous Office Action. The modification is necessitated by claim amendments.
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-6, 8-15, 17-20 and 22-23 are rejected under 35 USC § 101 because the claims are directed to non-statutory subject matter.
Step 1: Process, Machine, Manufacture or Composition
Claims 1-6, 8-9 and 22-23 are directed to a process, here a "computer-implemented method," with process steps “providing”, and “training”.
Claims 10-15 and 17-18 are directed to a machine or manufacture, here a “system”, with structural components like “one or more processors”, and “at least one memory”.
Claims 19-20 are directed to another machine or manufacture, here “one or more non-transitory computer-readable storage media”, with structural components like “one or more non-transitory computer readable storage medium”.
Step 2A Prong One: Identification of judicial exceptions
Claims 1, 10 and 19 recite:
Extract a plurality of frequent k-mers from the genomic sequence.
This step recites extracting k-mers out of the genomic sequence, which is a sequence manipulation that can be achieved in the human mind, with perhaps the aid of a pen and paper. Therefore, this step equates to an abstract idea of mental processes.
A defined subsequence of the first genomic sequence that has a character length of k an occurrence count within the genomic sequence that achieves a frequency threshold corresponding to the defined subsequence.
This step recites a mathematical calculation (count and comparing to a threshold), hence this step equates to an abstract idea of mathematical concepts.
Wherein an output value, of the plurality of output values, for the frequent k-mer, of the plurality of frequent k-mers, describes a likelihood that the frequent k-mer is a viral replication origin k-mer for the genomic sequence.
This step recites the correlation (here the likelihood) between the output value of frequent k-mer and the viral replication origin, which equates to the law of nature.
Storing, by the one or more processors, the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer.
This step recites a data storing operation in a general way. The k-mer, k-mer length and the frequency value of the k-mer, can easily be remembered in the human mind, or recorded on a paper by a human being. Therefore, this step equates to an abstract idea of mental processes.
The recitation “the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer” is directed to a mathematical concept because it recites number comparison (based on (a) the output value and (b) and a character length).
Hence, claims 1-6, 8-15, 17-20 and 22-23 do recite limitations that fall under the “mental processes” and “mathematical concepts” groupings of abstract ideas. Dependent claims provide further limitations to describe the k-mers and the neural network model, which are also fall into abstract ideas. The claims hence need be further analyzed in next step.
Step 2A Prong Two: Consideration of Practical Application
The claims result in a process to store the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set. The claims do not recite any additional elements that integrate the abstract idea/judicial exception into a practical application.
This judicial exception is not integrated into a practical application because the claims do not meet any of the following criteria:
An additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field;
an additional element that applies or uses a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition;
an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim;
an additional element effects a transformation or reduction of a particular article to a different state or thing; and
an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than
a drafting effort designed to monopolize the exception.
Step 2B: Consideration of Additional Elements and Significantly More
The claimed method also recites "additional elements" that are not limitations drawn to an abstract idea. The recited additional elements are drawn to:
Providing, by one or more processors, a first genomic sequence of a first genomic sequence set to a frequency-based k-mer extraction layer of a viral genome processing machine learning model to generate a plurality of frequent k-mers of the first genomic sequence (Claims 1, 10 and 19).
Providing, by the one or more processors, the plurality of frequent k-mers to a one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10 and 19).
To receive a plurality of output values respectively corresponding to the plurality of frequent k-mers, wherein an output value, of the plurality of output values, for the frequent k-mer, of the plurality of frequent k-mers, describes a likelihood that the frequent k-mer is a viral replication origin k-mer for the genomic sequence (Claims 1, 10 and 19).
One or more processors (Claim 10);
At least one memory (Claim 10);
One or more non-transitory computer readable storage media (claim 19).
A frequency-based k-mer extraction layer of a viral genome processing machine learning model (Claims 1, 10 and 19);
A one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10, 19 and 22); and
A two-dimensional convolutional neural network layer of a bacterial genome processing machine learning model (claims 6, 15).
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the first three additional elements are insignificant extra-solution activities of data gathering/data outputting (MPEP §2106.05(g)); the middle three elements are generic computing components that are used routinely in pertinent industries; and the last three are generic machine learning models that serve a technical environment.
The claims do not include additional elements that are sufficient to amount of significantly more than the judicial exception because it is routine and conventional to perform the acts of data analysis and generating new data. Other elements of the method include computing components which are recitations of generic computer structure and generic machine learning models/layers that serve to perform generic computing functions that are well-understood, routine, and conventional activities previously known to the pertinent industry. Viewed as a whole, these additional claim element(s) do not provide meaningful limitation(s) to transform the abstract idea recited in the instantly presented claims into a patent eligible application of the abstract idea such that the claim(s) amounts to significantly more than the abstract idea itself.
The core elements such as k-mers, and CNN are all well-known in the industry. Viral replication origin and antibacterial resistant genes are both explored before. The conventionality of the combined additional elements is witnessed in the following references:
Tampuu, et al. ("ViraMiner: Deep learning on raw DNA sequences for identifying viral genomes in human samples." PloS one 14.9 (2019): e0222271. Newly cited);
Cruz-Cano et al. ("Least-squares support vector machine approach to viral replication origin prediction." INFORMS journal on computing 22.3 (2010): 457-470. Newly cited).
Arango-Argoty et al. ("DeepARG: a deep learning approach for predicting antibiotic resistance genes from metagenomic data." Microbiome 6.1 (2018): 23. Newly cited); and
Guo et al. ("Seq2DFunc: 2-dimensional convolutional neural network on graph representation of synthetic sequences from massive-throughput assay." bioRxiv (2019): 2019-12. Newly cited).
Tampuu disclosed training datasets from NGS sequencing (@[38]) and machine learning models for classifying metagenomic datasets which are based on k-mer counting (@[24, 26, 27]). Cruz-Cano disclosed a support vector machine (SVM) approach to viral replication origin prediction. Arango-Argoty discloses DeepARG, a deep learning approach for predicting antibiotic resistance genes from metagenomic data (page 1, Title and Abstract). Guo disclosed Seq2DFunc, a 2-D CNN model for graph representation of synthetic sequences from massive-throughput assay.
Therefore, the 101 rejection is maintained.
Response to Applicant’s Arguments
In the Remarks filed 3/10/2026, Applicant argued (page 11, 2nd para through page 14, 2nd para) that “claim 1 presents patent eligible subject matter under at least both prongs of Step 2A of the Alice/Mayo test.”
In response, Applicant’s argument is not persuasive. At Step 2A/Prong one, claim 1 does recite “storing, by the one or more processors, the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer”, which is interpreted as a data storing operation in a general way. The k-mer, k-mer length and the frequency value of the k-mer, can easily be remembered in the human mind, or recorded on a paper by a human being. Therefore, this step of claim 1 equates to an abstract idea of mental processes. This step is also the ending step of independent claims 1, 10 and 19.
When all the additional elements are considered, individually or as a whole, the claims do not recite any additional elements that integrate the abstract idea/judicial exception into a practical application, because the claims do not meet any of the following criteria:
An additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field;
an additional element that applies or uses a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition;
an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim;
an additional element effects a transformation or reduction of a particular article to a different state or thing; and
an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than
a drafting effort designed to monopolize the exception.
Hence, At Step 2A/Prong one, claim 1 recites an abstract idea of mental processes; at Step 2A/Prong two, claim 1 is not integrated into a practical application.
The argument (page 14, 2nd para) that “claim 1 recites a combination of additional elements that improves a technical field such that the claim as a whole integrates any alleged abstract idea into a practical application” is not persuasive.
The recited additional elements are drawn to:
Providing, by one or more processors, a first genomic sequence of a first genomic sequence set to a frequency-based k-mer extraction layer of a viral genome processing machine learning model to generate a plurality of frequent k-mers of the first genomic sequence (Claims 1, 10 and 19).
Providing, by the one or more processors, the plurality of frequent k-mers to a one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10 and 19).
To receive a plurality of output values respectively corresponding to the plurality of frequent k-mers, wherein an output value, of the plurality of output values, for the frequent k-mer, of the plurality of frequent k-mers, describes a likelihood that the frequent k-mer is a viral replication origin k-mer for the genomic sequence (Claims 1, 10 and 19).
One or more processors (Claim 10);
At least one memory (Claim 10);
One or more non-transitory computer readable storage media (claim 19).
A frequency-based k-mer extraction layer of a viral genome processing machine learning model (Claims 1, 10 and 19);
A one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10, 19 and 22); and
A two-dimensional convolutional neural network layer of a bacterial genome processing machine learning model (claims 6, 15).
When these additional elements are considered, individually or as a whole, there is no technical improvement to the functioning of a computer (“machine learning technology” and “storage functionality”, page 13, 1st para). Appellant did not provide evidence the claimed computer system (memory, processors, and CNN layers) is anything more than a generic computer that implements the machine learning algorithm, etc., in the conventional and well- known way that computers ordinary function. This is not an improvement to the computer on which the program is run, but rather directed to an argument as to why such a program might be useful. Since the instant rejection is based on a finding that the claimed subject matter is directed to non-statutory subject matter, an argument that the claims satisfy the utility requirement under 101 misses point of the rejection and therefor does not address the actual basis of the rejection at hand.
In the remarks, Applicant argued (page 14, 4th para) that “the combination of additional elements of claim 1 are not "conventional" as alleged by the Office Action. Notably, the claims recite an unconventional machine learning architecture that provides non-routine results in the field of machine learning -and thus the claims provide an inventive concept.”
In response, Applicant’s argument is not persuasive. The argued “unconventional machine learning architecture” relates to the following steps in claim 1 (emphasis added):
Providing, by one or more processors, a first genomic sequence of a first genomic sequence set to a frequency-based k-mer extraction layer of a viral genome processing machine learning model to generate a plurality of frequent k-mers of the first genomic sequence (Claims 1, 10 and 19).
Providing, by the one or more processors, the plurality of frequent k-mers to a one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10 and 19).
To receive a plurality of output values respectively corresponding to the plurality of frequent k-mers, wherein an output value, of the plurality of output values, for the frequent k-mer, of the plurality of frequent k-mers, describes a likelihood that the frequent k-mer is a viral replication origin k-mer for the genomic sequence (Claims 1, 10 and 19).
More specifically refers to the following additional elements:
A frequency-based k-mer extraction layer of a viral genome processing machine learning model (Claims 1, 10 and 19);
A one-dimensional convolutional neural network layer of the viral genome processing machine learning model (Claims 1, 10, 19 and 22); and
As explained above, the computer and the one-dimensional CNN are at best the equivalent of merely adding the words “apply it” to the judicial exception. The “providing” and “to receive” were considered insignificant extra solution activity. This conclusion is re-evaluated in Step 2B. The limitations are mere data gathering and output recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II. The limitations remain insignificant extra-solution activity even upon reconsideration.
Even when considered in combination, the additional elements represent mere instructions to apply an exception and insignificant extra-solution activity, which cannot provide an inventive concept.
In the remarks, Applicant argued (page 14, penultimate para) that “the independent claims 10 and 19 recited patent eligible subject matter”.
In response, Applicant’s argument is not persuasive. independent claims 10 and 19 are very similar to independent claim 1. As discussed above, at Step 2A/Prong one, Step 2A/Prong two and Step 2B, the 35 USC 101 analytical conclusion holds the same.
Therefore, the 101 rejection is maintained.
Claim Rejections - 35 USC § 103
This is a maintained rejection. Modifications are necessitated by claim amendments.
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 1-5, 10-14, 19-20 and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Tampuu, et al. ("ViraMiner: Deep learning on raw DNA sequences for identifying viral genomes in human samples." PloS one 14.9 (2019): e0222271. Previously cited), and further in view of Cruz-Cano et al. ("Least-squares support vector machine approach to viral replication origin prediction." INFORMS journal on computing 22.3 (2010): 457-470. Previously cited), and Liu et al. ("Estimation of genomic characteristics by analyzing k-mer frequency in de novo genome projects." arXiv preprint arXiv:1308.2012 (2013). Newly cited).
Claim 1 is interpreted as a method to use k-mer and k-mer frequency from viral genome sequence for machine learning prediction. Regarding claim 1,
Tampuu provides (page 1, section “Abstract”) ”we developed ViraMiner, a deep learning-based method to identify viruses in various human biospecimens”, which suggests a computer implemented method.
Tampuu provides (page 11, 2nd para) ”the training dataset included 19 different NGS experiments that were analyzed and labeled by PCJ-BLAST [38] after applying the de novo genome assembly algorithms”, which teaches acquiring the first genome dataset.
Tampuu provides (page 4, 2nd para) “most of the previous machine learning models for classifying metagenomic datasets are based on k-mer counting”, which teaches k-mer extraction/counting as known for metagenomic classification.
Tampuu provides (page 11, last para) ”The convolutional layers treat their inputs as multidimensional arrays—an image is treated as a 2D array with 3 color channels, while a DNA sequence is seen as a 1D array with one channel per possible nucleotide value. In the present work we use sequences of length 300 and consider 5 possible values (ATGCN) at each position, corresponding to a 1D sequence of length 300 with 5 channels”, which teaches the 1D-CNN layer for DNA sequence processing.
Tampuu provides (page 1, section “Abstract”) ”pattern-frequencies on raw metagenomics contigs” and (page 4, 2nd para) “most of the previous machine learning models for classifying metagenomic datasets are based on k-mer counting”, which teaches frequency-based sequence feature extraction.
Tampuu does not teaches a threshold for the k-mer frequence scores.
Tampuu provides (page 1, section “Abstract”) ”ViraMiner contains two branches of Convolutional Neural Networks designed to detect both patterns and pattern-frequencies on raw metagenomics contigs”, which teaches CNN processing of sequence pattens.
Tampuu provides (page 4, 1st para last three lines) “Sigmoid activation function is applied in this node, transforming the weighted sum of inputs to a probability (into range [0, 1])”, which teaches probability/likelihood output from CNN model.
Tampuu does not teach viral replication origin.
Cruz-Cano disclosed a support vector machine approach to viral replication origin prediction (page 1, Title and Section “Abstract”: “the LS-SVM approach further improves the prediction accuracy for the herpesvirus replication origins”).
Tampuu provides (page 5, penultimate para) “the best test performance with RF models was achieved with 6-mers and it produced test AUROC 0.875 (Fig 4). RF performances on 3- 4- 5- and 7-mers were 0.867, 0.872, 0.873, and 0.869 respectively”, which teaches k-mer length variation and Cruz-Cano provides (page 1, Title and Section “Abstract”) “the LS-SVM approach further improves the prediction accuracy for the herpesvirus replication origins”, which teaches viral replication origins. Storing selected candidates based on score and length is likely an obvious implementation detail. Storing a hash by the sorted keys (here the k-mers are the hash keys, wherein the lengths of the k-mers are automatically sorted using the “sort” function that usually come with the Hash data structure) is a basic hash function (in pseudo code this is probably implemented by something like “for each sorted k-mer, print/store “k-mer\t Frequency{“k-mer”}).
Cruz-Cano does not teaches a threshold for the k-mer frequence scores neither.
Liu teaches thresholds for k-mer frequency (page 9, last para lines 2-3; page 13, last para lines 3-4; page 43 penultimate para last two lines; page 45 2nd para lines 6-8).
Regarding claim 2, Cruz-Cano provides (page 2, 2nd para) “Like ANN, support vector machines (SVMs) can learn from characteristics of the known replication origins of those genomes in the training data set and then make predictions of where the replication origins of a new genome are likely to be”, which teaches generating training data with known replication origins through a machine learning model.
Early studies on the genome.
Cruz-Cano provides “DNA sequences of herpesviruses have suggested that replication origins often lie around regions with unusually high concentration of palindromes, where a palindrome is a stretch of DNA bases followed immediately by its reverse complement. Based on these observations, Leung et al. (2005) suggest a computational method using the scan statistics to locate significant clusters of palindromes and predict the likely locations of replication origins” (page 2, 2nd para), which teaches clustering viral genomes by the replication origin k-mer because a palindrome is a k-mer.
Regarding claim 3, Cruz-Cano provides “we apply SVM technique for the replication origin prediction. Furthermore, we include more genome features, namely, family/subfamily information and the A+T content (recently shown in Chew et al. (2007) to be an important feature for prediction) as our input variables for SVMs” (page 2, 4th para), which suggests measuring viral genome similarities and predict the viral strain to a family/subfamily.
Regarding claim 4, Cruz-Cano provides “K-mer models. Most of the previous machine learning models for classifying metagenomic datasets are based on k-mer counting [24, 26, 27]. To compare against the performance of such methods, we also extracted k-mers from the investigated dataset and trained Random Forest (RF) classifiers on the extracted values, while keeping the same data partitioning as above” (page 5, 2nd para), which teaches extracting k-mers and counting the extracted k-mers.
Neither Cruz-Cano nor Tampuu teach a k-mer threshold. Liu teaches thresholds for k-mer frequency (page 9, last para lines 2-3; page 13, last para lines 3-4; page 43 penultimate para last two lines; page 45 2nd para lines 6-8).
It is obvious to convert the counts to scores (such as 0 to 100 or 0 to 1) which will allow a universal threshold score for determining the frequent k-mers.
Regarding Claim 5, neither Cruz-Cano nor Tampuu teach a k-mer threshold. Liu teaches thresholds for k-mer frequency (page 9, last para lines 2-3; page 13, last para lines 3-4; page 43 penultimate para last two lines; page 45 2nd para lines 6-8).
It is obvious to set up an arbitrary threshold score so the k-mer with a frequency score above the threshold score is defined as frequent k-mer.
Since Cruz-Cano provides “the programs used to determine the performances of the LS-SVMs were run in a Dell Optiplex 745 Minitower, PentiumD 945/3.40GHz using MATLAB R2007 with Windows XP” (page 10, 2nd para), which teaches a computer system, the art applied to claims 1-5 also teaches claims 10-14.
Similarly, the art applied to claims 1-2 also reads on claims 19-20.
Regarding claim 22, Tampuu provides “Therefore, the models in the Results section are obtained with step-wise training procedure, rather than end-to-end or fine-tuning” (page 13, 4th para) and Fig. 6 (page 12, Fig. 6: “with best performing layer sizes shown”), which teaches the one-dimensional neural network layer is selected (to reach the “best performing layer sizes”) based on the associated kernel size (listed in the Fig. 6b and 6c)
Regarding claim 23, Tampuu provides (page 5, penultimate para) “the best test performance with RF models was achieved with 6-mers and it produced test AUROC 0.875 (Fig 4). RF performances on 3- 4- 5- and 7-mers were 0.867, 0.872, 0.873, and 0.869 respectively”, which teaches comparing k-mer lengths. Preference for shorter k-mers is not directly taught; likely an optimization/design choice if shorter candidate origins reduce redundancy or improve generalization.
It would have been prima facie obvious to modify Tampuu’s viral genome sequence processing pipeline which uses k-mer frequency and a CNN to identify the existence of viral genome in human samples, with Cruz-Cano’s method of identifying viral replication origin. Replication of DNA genomes is a central step in the reproduction of many viruses. Because “procedures to find replication origins, which are initiation sites of the DNA replication process, are therefore of great importance for controlling the growth and spread of such viruses” (Cruz-Cano: page 1, Section “Abstract”). Predicting the viral replication origin includes predicting the “existence of virus” and would serve additional purposes to predicting the existence of a virus.
One would reasonably expect success as the same input viral genome sequences are available to both Tampuu and Cruz-Cano for model training and Cruz-Cano teaches that “first, the method uses only a single sequence feature, namely the palindrome distribution, of the genome and does not offer any obvious generalization for multiple sequence features to be simultaneously taken into consideration. Second, when predicting replication origins for one herpesvirus, relevant information from other members of the viral family is not used. In (Cruz-Cano et al., 2007), we have proposed to address these issues by a machine learning approach, namely artificial neural networks (ANN). Like ANN, support vector machines (SVMs) can learn from characteristics of the known replication origins of those genomes in the training data set and then make predictions of where the replication origins of a new genome are likely to be” (Cruz-Cano: page 2, 2nd para), one have a good reason to expect Tampuu’s CNN model will perform well in predicting viral replication origin too. Because Cruz-Cano’s palindrome is a k-mer, under a BRI; and Cruz-Cano’s “palindrome distribution” encompasses the k-mer frequency. Tampuu’s CNN model has a “Pattern Branch” and a “Frequency Branch” (Tampuu :page 3, Fig. 1; page 12, Fig. 6) which are ready to hand the “palindrome distribution” in the form of k-mer and k-mer frequency in model training.
It would have been prima facie obvious to combine the viral genome sequence processing pipeline of Tampuu and Cruz-Cano which uses k-mer frequency and a CNN to identify the viral replication origin (hence identify the existence of virus), with Liu’ method that apply thresholds in k-mer frequency counting. Because Liu’s k-mer frequency analysis with thresholding can be used as a general and assembly-independent method for estimating genomic characteristics, which can improve our understanding of a species genome (Liu: pages 1-2, Section “Abstract/Conclusion”).
One would reasonably expect success as Tampuu, Cruz-Cano and Liu are all about genomic sequence study. Tampuu’s CNN model applied k-mers and k-mer frequency directly, Cruz-Cano’s palindrome is a k-mer, and Cruz-Cano’s “palindrome distribution” encompasses the k-mer frequency. Combining Liu’s k-mer frequency thresholding would be:
“Combining prior art elements according to known methods to yield predictable results” (MPEP 2143.III.(A)).
“Combining prior art elements according to known methods to yield predictable results” (MPEP 2143.III.(A)).
Claims 6 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Tampuu, Cruz-Cano and Liu, as applied to claims 1-5, 10-14, 19-20 and 22-23 above, and further in view of Arango-Argoty et al. ("DeepARG: a deep learning approach for predicting antibiotic resistance genes from metagenomic data." Microbiome 6.1 (2018): 23. Previously cited) and Guo et al. ("Seq2DFunc: 2-dimensional convolutional neural network on graph representation of synthetic sequences from massive-throughput assay." bioRxiv (2019): 2019-12. Previously cited).
Claim 6 is interpreted as method to predict resistant bacterial segment from bacteria genome using deep learning. Regarding claim 6, Arango-Argoty discloses DeepARG, a deep learning approach for predicting antibiotic resistance genes from metagenomic data (page 1, Title and Abstract). Particularly,
Arango-Argoty provides “in the study, a large scale analysis was carried out by screening thousands of metagenomes and bacterial genomes to a curated set of beta lactamases” (page 9, col 1, last para), which teaches identifying bacteria genome sequence set (e. g. “a second genomic sequence set” that is different from the 1st genomic sequence set of virus).
Arango-Argoty provides “evaluation of the deep learning models over 30 antibiotic resistance categories demonstrates that the DeepARG models can predict ARGs with both high precision (> 0.97) and recall (> 0.90)” (page 1, section “Abstract/Results”), which teaches determining resistant bacterial segments, because a gene is a sequence segment.
Arango-Argoty does not teach 2-dimensional convolutional neural network. Guo disclosed Seq2DFunc, a method to predict sequence function using 2-dimensional convolutional neural network (page 1, Title and Abstract).
It would be obvious once the ARG (antibiotic resistance gene) is identified, to recommend an alternative antibiotic that will not be targeted by the ARG protein.
Since Cruz-Cano provides “the programs used to determine the performances of the LS-SVMs were run in a Dell Optiplex 745 Minitower, PentiumD 945/3.40GHz using MATLAB R2007 with Windows XP” (page 10, 2nd para), which teaches a computer system, the art applied to claim 6 also teaches claim 15.
It would have been prima facie obvious to combine the combined Tampuu, Cruz-Cano, and Liu pipeline which uses k-mer frequency and a CNN to identify viral replication origin (hence the existence of viral infection) in human samples, with Arango-Argoty’s DeepARG which predicts antibiotic resistance genes from metagenomic data using a deep learning approach (page 1, Title and Abstract), because when a patient walks into a hospital with infections, the clinical lab would be obvious to check possible bacterial infection after examining viral infection.
One would reasonably expect success because after acquiring sequencing data from the patient samples, identify viral replication origin using k-mer-based one-dimensional CNN layer; and predicting antibiotic resistance genes from metagenomic data using DeepARG (a CNN model) (Arango-Argoty: page 1, Title and Abstract), can work independently without interfering each other in one computer. The would be classic example of:
“Combining prior art elements according to known methods to yield predictable results” (MPEP §2141.III.(A)).
It would have been prima facie obvious to modified the combined Tampuu, Cruz-Cano, Liu and Arango-Argoty pipeline which uses k-mer frequency and a CNN to identify viral replication origin (hence the existence of viral infection) in human samples, as well as use the DeepARG to predict antibiotic resistance genes from metagenomic data (Arango-Argoty: page 1, Title and Abstract), with Guo’s Seq2DFunc pipeline which predict sequence function using a 2D convolutional neural network on graph representation of sequences (Guo: page 1, Title and Abstract). Because Guo’s deep learning approach trains a two-dimensional convolutional neural network (CNN) on an ordered graph representation of nucleic acid sequences to predict their functions (Seq2DFunc) (page 1, Section Abstract).
One would reasonably expect success because both Arango-Argoty’s DeepARG and Guo’s Seq2DFunc both use deep learning approach for modeling nucleic sequences, and Guo demonstrated superior performance by the 2D CNN vs the 1D CNN (Guo: page 1, Section Abstract). The would be classic example of:
“Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention” (MPEP §2141.III.(G)).
Claims 8-9 and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over Tampuu, Cruz-Cano, Liu, Arango-Argoty and Guo, as applied to claims 1-6, 10-15, 19-20 and 22-23 above, and further in view of Baker et al. ("Genomic insights into the emergence and spread of antimicrobial-resistant bacterial pathogens." Science
360.6390 (2018): 733-738. Previously cited).
Regarding claims 8-9, neither Arango-Argoty nor Guo teaches a location distribution or the demographic distribution of identified ARG sequence segments. Baker provides Fig. 1 (page 2, Fig. 1), which shows the time line of the ST22 MRSA (methicillin resistance) pandemic in a Bayesian phylogenetic tree in 14 countries.
Baker teaches a location-wide designation and the demographic distribution of identified ARG sequence segments (here MRSA), because the countries represent locations, and demographic populations.
The art applied to claims 8-9 also teaches claims 17-18.
The combined Tampuu, Cruz-Cano, Liu, Arango-Argoty and Guo predicted viral replication origin and analyzed antimicrobial resistant sequence segment successfully using CNN models. Baker, on the other hand, profiled and showed the location and demographic distributions of MRSA with a vivid graph (page 2, Fig. 1). Baker’s method would allow the combined Tampuu, Cruz-Cano, Liu, Arango-Argoty and Guo to present their identified drug resistance sequence segments in an informative and impressive way. This is a classic example of: “Combining prior art elements according to known methods to yield predictable results” (MPEP §2141.III.(A)).
Response to Applicant’s Argument
In the Remarks filed 3/10/2026, Applicant argued (page 16, 2nd para. Emphasis addd) that “Tampuu does not describe the use of a frequency threshold to extract a plurality of frequent k-mers, or a frequency threshold that corresponds to the defined subsequence, as recited by the claims. Tampuu is also silent with respect to providing each of the plurality of frequent k-mers to a one-dimensional convolutional neural network layer to receive a plurality of output values. None of the other cited references cure at least this deficiency with Tampuu”.
In response, Applicant’s argument is not persuasive. Tampuu provides (page 4, 2nd para) “most of the previous machine learning models for classifying metagenomic datasets are based on k-mer counting”, which teaches k-mer extraction/counting as known for metagenomic classification.
Tampuu provides (page 1, section “Abstract”) ”pattern-frequencies on raw metagenomics contigs” and (page 4, 2nd para) “most of the previous machine learning models for classifying metagenomic datasets are based on k-mer counting”, which teaches frequency-based sequence feature extraction. Liu teaches thresholds for k-mer frequency (page 9, last para lines 2-3; page 13, last para lines 3-4; page 43 penultimate para last two lines; page 45 2nd para lines 6-8) multiple times, directly.
Tampuu provides (page 11, last para) ”The convolutional layers treat their inputs as multidimensional arrays—an image is treated as a 2D array with 3 color channels, while a DNA sequence is seen as a 1D array with one channel per possible nucleotide value. In the present work we use sequences of length 300 and consider 5 possible values (ATGCN) at each position, corresponding to a 1D sequence of length 300 with 5 channels”, which teaches the 1D-CNN layer for DNA sequence processing.
Liu teaches thresholding for k-mer frequency (page 9, last para lines 2-3; page 13, last para lines 3-4; page 43 penultimate para last two lines; page 45 2nd para lines 6-8).
In the Remarks filed 3/10/2026, Applicant argued (page 16, 3rd para) that “Cruz-Cano does not describe providing each of a plurality of frequent k-mers to a one-dimensional convolutional neural network layer to receive a plurality of output values, nor does the reference describe storing the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer, as recited by claim 1. None of the other cited references cure at least this deficiency with Cruz-Cano”.
In response, Applicant’s argument is not persuasive. Tampuu teaches CNN processing of sequence pattens, more specifically the 1D-CNN layer for DNA sequence processing, as discussed above.
Cruz-Cano provides (page 1, Title and Section “Abstract”) “the LS-SVM approach further improves the prediction accuracy for the herpesvirus replication origins”, which teaches prediction of viral replication origins. Storing selected candidates based on score and length is an obvious implementation detail.
Cruz-Cano provides “DNA sequences of herpesviruses have suggested that replication origins often lie around regions with unusually high concentration of palindromes, where a palindrome is a stretch of DNA bases followed immediately by its reverse complement. Based on these observations, Leung et al. (2005) suggest a computational method using the scan statistics to locate significant clusters of palindromes and predict the likely locations of replication origins” (page 2, 2nd para), which teaches clustering viral genomes by the replication origin k-mer because a palindrome is a k-mer.
Because Cruz-Cano’s palindrome is a k-mer, under a BRI; and Cruz-Cano’s “palindrome distribution” encompasses the k-mer frequency. Further, character length is an internal part of the palindrome. Tampuu’s CNN model has a “Pattern Branch” and a “Frequency Branch” (Tampuu :page 3, Fig. 1; page 12, Fig. 6) which are ready to hand the “palindrome distribution” in the form of k-mer and k-mer frequency in model training.
Therefore, it is obvious that the combined Cruz-Cano and Tampuu input frequent k-mers to a 1D-CNN layer to receive a prediction output, nor does the reference describe storing the frequent k-mer as one of a plurality of viral replication origin k-mers for the genomic sequence set based on (a) the output value and (b) and a character length of the frequent k-mer, as recited by claim 1. Storing a hash by the sorted keys (here the k-mers are the hash keys, wherein the lengths of the k-mers are automatically sorted using the “sort” function that usually come with the Hash data structure) is a native hash function (in pseudo code this is probably implemented by “for each sorted k-mer, print/store “k-mer\t Frequency{“k-mer”}).
Therefore, the 103 rejection is maintained.
Additional Noted Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Chan, Helen HW, et al. "{HashKV}: Enabling efficient updates in {KV} storage via hashing." 2018 USENIX Annual Technical Conference (USENIX ATC 18). 2018.
This article described Hash data structure (e. g. key-value pairs indexable by the keys) and the sort functions appliable to the Hash data structure. The hash function itself has nothing to do with the replication origin identification.
Conclusion
No claims are allowed.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GUOZHEN LIU whose telephone number is (571)272-0224. The examiner can normally be reached Monday-Friday 8-5.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Larry D Riggs can be reached at (571) 270-3062. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/GL/
Patent Examiner
Art Unit 1686
/Anna Skibinsky/
Primary Examiner, AU 1635
Prosecution Insights